Edinburgh Napier University

  For assembly purposes, robot accuracy is critical only in small well-defined volumes, such as immediately over a pallet or work table where pieces are being lifted and assembled. Since the volumes are small it is possible to consider the robot system as a black box, with the target pose (position and orientation) as inputs and the inaccuracies as outputs. A factorial experimental design was developed and conducted to systematically collect data. The data has been used to generate models which explain the variance of the inaccuracies. The factorial experiment involved collecting 864 data points. To reduce the time required to take all the necessary measurements, smaller sets of data can be used to develop simpler regression models for each work volume. The explanatory power of alternative regression models based on reduced data sets has been examined.